Ben Talbot

creating-montage-videos-with-ffmpeg

Short post describing how to programmatically create a video montage from a collection of videos and photos. Jotting this down, so at the very least it'll help future me with FFMPEG.


Creating montage videos with FFMPEG

removing-blank-pdf-pages

Cleaning out my filing cabinet meant scanning 1000s of pages of PDFs, with 100s of blank pages littered throughout. Here's how I chained together some basic Linux tools to get rid of them.


Removing 100s of blank pages from PDFs

creating-hibernate-shortcut-ubuntu

Getting the power button to hibernate with GNOME in Ubuntu 20.04 proved troublesome. Here's a workaround to hibernate with a key press.


Creating a hibernate shortcut in Ubuntu 20.04

getting-hibernate-working-ubuntu

Recently hibernate stopped working on my Ubuntu 18.04 laptop. After spending an age last time trying to get it to work again, I thought I'd document what ended up working this time.


Getting hibernate to work in Ubuntu 18.04+

Feras Dayoub

Ruth Schulz

Ben Upcroft

Peter Corke

Gordon Wyeth

View the code on GitHub

IEEE TCDS journal

Brisbane Times article

The LABS article (p. 88)

PhD thesis

ICRA2016 paper

ACRA2015 paper

RSS2015 workshop paper

ICRA2015 paper

ARC Discovery Project - DP140103216

QUT Centre for Robotics (QCR)

abstract-map

The Abstract Map allows a robot to navigate in unseen spaces using navigation symbols. It was developed throughout my PhD and produced five research publications, a doctoral thesis, and four different software implementations.


Began PhD, as part of the Human Cues for Robot Navigation Australian Research Council (ARC) Discovery project

First version of the abstract map presented as part of a floor plan-based navigation system at ICRA 2015

Concept of an abstract map built from spatial descriptions presented to the Workshop on Model Learning for Human-Robot Communication at RSS 2015

Navigation using an abstract map built from natural language phrases presented at ACRA 2015, and won best student paper award

Demonstrated the abstract map navigating a robot through real built environments for ICRA 2016 paper

Thesis published on integrating symbolic spatial information in robot navigation, with the abstract map as the core contribution

Code for the final version of the abstract map publicly released (v1.1.0)

Navigation studies comparing abstract map-based robot navigation with human performance in real environments published in IEE TCDS, as the conclusion of this work

Research results reported nationally, including coverage by Fairfax media

The Abstract Map

David Hall

Haoyang Zhang

Suman Bista

Rohan Smith

Niko Sünderhauf

BenchBot blog for NVIDIA

arXiv BenchBot paper

IJRR BEAR paper

Go to the tutorial

EvalAI challenge

Overview of the scene understanding project

NVIDIA

Australian Centre for Robotic Vision (ACRV)

benchbot

BenchBot is a software suite for benchmarking  robotics research across both 3D simulation and real robot platforms. The software provides a simple Python interface to the robot's underlying sensorimotor capabilities.


First internal presentation of BenchBot concept to ACRV review panel, focusing only on real robots

Internal alpha release of BenchBot as a bridge between simulation and real robots (v0.1.0)

First mainstream release, ready for public use (v1.0.0)

Preprint released on arXiv describing the BenchBot software suite

Support for more research problems, simulators, and robot platforms, through a new modular architecture powered by add-ons (v2.0.0)

Robotic Vision Scene Understanding (RVSU) challenge released, powered by BenchBot, as part of the Embodied AI workshop at CVPR 2021

Robotic Vision Scene Understanding (RVSU) challenge released, powered by BenchBot, as part of the Embodied AI workshop at CVPR 2022

First version released with support for NVIDIA's Omniverse-powered Isaac Sim (v2.2.0)

BenchBot environments for active robotics (BEAR) published in IJRR

Technical blog featuring BenchBot released on NVIDIA's developer website

BenchBot Software Stack

Gavin Suddrey

Steven Martin

Getting started tutorial

Problem solving with trees

behaviour-trees

A behaviour trees library to make complex robotics problems feasible. Features include: automatic data transfer between leaves, configurable leaf configurations with basic lambda functions, and a powerful leaf lifecycle.


Library commenced, initially as a tool for managing behaviours on our mobile manipulation robot

Used as the system orchestrator in our entry for the 2019 RoboCup@Home competition

Began generalising library as a standalone solution for powering complex robotic systems with behaviour trees

Released internally, version used on robotic systems from arms to mobile bases, and everything in between (v0.1.0)

First official open source release (v1.0.1)

Updated list of robotics problems library has been applied to (v1.1.0)

Behaviour Trees for Robots

Read the paper

View publication on IEEE

View preprint on arXiv

Watch the robot navigate

abstract-map-2020-tcds

Human navigation in built environments depends on symbolic spatial information which has unrealised potential to enhance robot navigation capabilities. Information sources such as labels, signs, maps, planners, spoken directions, and navigational gestures communicate a wealth of spatial information to the navigators of built environments; a wealth of information that robots typically ignore. We present a robot navigation system that uses the same symbolic spatial information employed by humans to purposefully navigate in unseen built environments with a level of performance comparable to humans.


Robot Navigation in Unseen Spaces using an Abstract Map

Visit Emobided AI workshop site

embodied-ai-2022-arxiv

We present a retrospective on the state of Embodied AI research. Our analysis focuses on 13 challenges presented at the Embodied AI Workshop at CVPR. These challenges are grouped into three themes: (1) visual navigation, (2) rearrangement, and (3) embodied vision-and-language. We discuss the dominant datasets within each theme, evaluation metrics for the challenges, and the performance of state-of-the-art models. We highlight commonalities between top approaches to the challenges and identify potential future directions for Embodied AI research.


Retrospectives on the Embodied AI Workshop

Frederic Mair

View the behaviour trees project

lingua-2022-ral

Domestic and service robots have the potential to transform industries such as health care and small-scale manufacturing, as well as the homes in which we live. However, due to the overwhelming variety of tasks these robots will be expected to complete, providing generic out-of-the-box solutions that meet the needs of every possible user is clearly intractable. To address this problem, robots must therefore not only be capable of learning how to complete novel tasks at run-time, but the solutions to these tasks must also be informed by the needs of the user. In this paper we demonstrate how behaviour trees, a well established control architecture in the fields of gaming and robotics, can be used in conjunction with natural language instruction to provide a robust and modular control architecture for instructing autonomous agents to learn and perform novel complex tasks.


Learning and Executing Re-usable Behaviour Trees from Natural Language Instruction

View the dataset through QCR

View the publication on Sage

bear-2022-ijrr

We present a platform to foster research in active scene understanding, consisting of high-fidelity simulated environments and a simple yet powerful API that controls a mobile robot in simulation and reality. In contrast to static, pre-recorded datasets that focus on the perception aspect of scene understanding, agency is a top priority in our work. We provide three levels of robot agency, allowing users to control a robot at varying levels of difficulty and realism.
